RE01 Rice Brian T. EM2
Sat Nov 10 07:18:01 2001


> >>>>> "RRBT" == RE01 Rice Brian T EM2 <> writes:
> RRBT> If you look at the Tunes CVS snapshots (best available as .bz2),
> I can't - can't figure out where to get it from.  could you help
> please. should have it somewhere, but my really bad
connection makes it impossible for me to locate it exactly.

> RRBT> you'll find some Scheme source code written as some prototype
> RRBT> building blocks for Tunes' semantics. I run a project
> Ah making things easy, hey?  Askemos has Scheme code between the
> comments as well.

Oh, good. But I still can't connect to your site's documentation: I'm being
blocked on that port on my end, and the administrators here are very stupid
and paranoid here, I'm sorry to say.

> RRBT> alongside/as-prototype-for Tunes called Arrow
> RRBT> ( However, because of my current
> RRBT> location, the code updates are lagged by a few months.
> Today I read deeper into the documents.  Interesting reads, but pretty
> demanding for a none native speaker after a working week.  The Arrow
> paper is beyond my capability of reception for today.

Fair enough. It's not a great paper anyway. I'll be re-writing it when I
have a decent document preparation system working again.

Basically Arrow is a system that models everything as Cons-cells, but lacks
Lisp's semantic model. Right now, it implements Relational Algebra lazily
(flawlessly on my own copy of the code, but the copy on the Tunes server is
outdated and buggy - still it is self-documenting). Relational algebra
relates quite well to concurrent equational membership rewrite logic
semantics (whew! :) that Maude has, though. Equational rewrite logic is just
the logic of term replacement in symbolic equations as an evaluation/proof
model. 'membership' refers to enrichment with a type system (very suitably
extensible by the language). Incidentally, Maude has been very interestingly
used to model and execute concurrent (and distributed) object-oriented
system semantics; and Tunes'ers should find it interesting that most of its
best features are due to reflective facilities. Basically I am building off
of all these ideas and Category Theory.

> RRBT> Basically, Francois-Rene Rideau (Fare) and I are doing active
> RRBT> research to see how to design Tunes' implementation, and David
> If nothing else, than at least we agree that todays OS's are
> fundamentally flawed.  Next there are some fundamental design
> decisions (separation of aspects: functional vs. dynamic description,
> orthogonality, persistence and importance of principles like security
> and liberty), which I see we share.

Perhaps. I think you agree with Tunes mostly, but I don't think in terms of
principles or separation of aspects per se (and in this way I differ from
many Tunes members). I have a design goal in mind, basically, and it happens
to match Tunes relatively well enough.

> RRBT> Manifold (Tril) manages the web site. Most others are watching
> RRBT> and waiting for a solid plan from us (am I wrong? someone
> RRBT> correct me on this, please).
> As for the web site I could imagine that you might like the layered
> Wiki implementation, which comes with Askemos.  It's unfortunately not
> yet documented (I'm willing upon demand), it allows to search for
> links in several Wiki data bases.  The search order can be altered
> depending on point of view.  Use example: I look at a term (say
> CustumerXY) from the technology view and see details, from the
> business view and see "big reference" and from the personal view and
> see "beware the boss".  Furthermore you can include pages into each
> other and have several views at it.

This sounds really good. Perhaps Tril can do something with this.

> One of the more fascinating moments: I just added a few notes about
> Tunes and found that it has a great context view (read well balanced)
> http://localhost:7180/A7784cc3f5c966c1e27a195b68070b87b/TUNES?
> template=context

Heh, you gave a localhost-relative URL. :) Still, I can't read it.

> I don't have enough usage experience yet.  A graphical view of the
> network between the pages would be great...

Sure. You should look sometime at a hypertext documentation system for the
Lisp Machine called Concordia. It's got such a feature, although it was
written before HTML existed, so it's not compatible with web systems.
Perhaps CL-HTTP has some feature for that, or at least could be used for it.

> >> Next I'd like to invite you to review As I
> >> said, the design goals where pretty simillar.  We would like to
> >> share ideas and code as much as possible - after all who want's to
> >> duplicate the work?
> RRBT> Unfortunately, from my location, only the front page is rendered
> RRBT> at all, which doesn't help me out much. I'd estimate that if you
> It doesn't help at all.  You can't access other ports than 80?  So I
> better don't ask what's you location is - you wan't be allowed to tell
> anyway ;-)

Actually you can see us on weather satellite photos. :P But you're right, I
can't say where we are.

> That's a pity.  You can get everything (code and docs) from

Ah! Thank you. Er... the file didn't transfer properly (we have a very high
latency factor in our satellite uplink). Please send it to me privately via

> But only the readme will be of use until you have build it.
> Beware that documents are in infancy.

Of course.

> RRBT> can compare your system's semantics (or its intended semantics)
> RRBT> with those of, say, a given Lisp like Scheme, or perhaps some
> RRBT> other language that supports really expressive semantics, then
> RRBT> we would have a very good idea how to compare the ideas. If
> May I first take off the XML-is-everything marketing envelope (Askemos
> is supposed to be useful, which includes market fit): S-expressions ==
> XML.


> Askemos defines a virtual machine at document level, wich works within
> an abstract infromation space.

Document level? Perhaps you can restate this as specification level,

> Orthogonality, pure concepts, functional programing, dynamic modelling
> using petri nets, persistence etc. etc.

Interesting. I really do suggest you look into Maude, despite it's non-SEXP
syntax. I know that it certainly has an algebraic (executable) definition of
Petri Nets, for example.

> Information is understood to be multidimentional.  Three axis so far:
> structure, context and rights.


> The most interesting success so far is a capability system, which
> doesn't need an administrative super power.  (Excluding a systematic
> error of todays OSes: whom's account to hack if there's no root -
> boring business.)  It's only a few lines of code.

I'll look.

> Next it doesn't trust a particular machine to operate correct.
> Reliability is achived by redundant execution of transactions and
> voting on the result.  This part is only prototyped at the moment.

> RRBT> you're interested in a direct review on the OS pages, you can
> RRBT> write to the Review subproject mailing list, but there's a long
> RRBT> waiting list for adding entries to those pages. :)
> Unfortunately I'm sort of forced to do other things to make my living.

As you can see, so am I. :)
However, I will be out of this job in late January, and I am applying to MIT
with a research project proposal. I also will be starting a small company,
whose function will be at least to allow me to write off research expenses
(which keep piling up no matter how economically I run things).

> Regards
> /Jerry